Managers Are Not Overhead – They’re the Infrastructure That Makes AI Work

When the post‑COVID tech layoffs started rolling in late 2022, a familiar refrain echoed through the halls of Meta, Google, Amazon and countless other firms: “let’s flatten the org, cut the bureaucracy.” The image was clear—fewer layers of management, slimmer reporting lines, a leaner, faster company. On paper it looked logical, especially as AI tools began to promise things that once sat squarely on a manager’s desk: drafting performance reviews, scheduling meetings, even nudging teams toward the right priorities.

Fast‑forward to mid‑2026 and the narrative has taken on a life of its own. Some venture‑backed startups are experimenting with one manager for fifty individual contributors, arguing that AI can handle most of the coordination work. The math seems simple: if a bot can generate a project plan, why do we need a human to approve it? If an algorithm can flag a risk, why does a supervisor need to raise the alarm?

But that line of thinking glosses over a growing body of evidence that managers are, in fact, becoming the very foundation—​the infrastructure—​that lets AI deliver value in the first place. The recent Microsoft 2026 Work Trend Index is a case in point. It found that cultural and managerial factors account for twice the impact of AI on employee performance compared with raw individual effort. In other words, it’s not just the algorithm that matters; it’s the people who shape how it’s used.

When managers actively model AI usage, their teams report a 17‑point jump in perceived AI value, a 22‑point rise in critical thinking about AI, and a 30‑point boost in trust toward agentic systems. When those same leaders build psychological safety around experimentation, employees are up to 20 points more ready to adopt AI and 1.4 × more likely to become high‑frequency users. The data is loud and clear: managers amplify the return on AI investments.

The effect is even sharper for the so‑called “Frontier Professionals”—the roughly 16 % of workers who string together multi‑step workflows and even build their own multi‑agent pipelines. Those folks are far more likely to have managers who themselves use AI (85 % vs. 64 % for the broader pool), set clear quality standards, champion experimentation, and reward redesign attempts regardless of outcome. Microsoft calls the tension between fast‑moving individuals and slower‑moving structures the “Transformation Paradox.” Their conclusion? Managers are the layer that resolves it, translating high‑level AI strategy into day‑to‑day practices that actually create value.

What about the claim that once AI is fully baked into a company’s DNA, the managerial role will shrink? If that were happening, we’d already see the warning signs. Instead, the LeadDev Engineering Leadership Report 2026—based on responses from 600 engineering leaders, more than half of whom manage teams—shows the opposite trend.

According to the survey, AI is expanding both the technical scope and the organizational expectations placed on leaders, without lightening their workload. Here’s what the numbers say:

• 63 % say their area of responsibility grew in the past year.
• 60 % report more communication with team members, customers, and stakeholders.
• 22 % now have additional teams reporting to them.
• 29 % have more direct reports.
• Architectural decisions and technical strategy claim the biggest share of their newly‑added time.

One could read these stats as confirmation that “flattening” is succeeding—more people per manager, bigger spans of control. A second, equally valid reading is that the role is still in flux: managers are juggling the old, ever‑expanding load while adding the new, AI‑centric responsibilities that come with overseeing intelligent automation. Either way, the evidence does not point to a contraction of the managerial function; it points to an evolution.

That’s not to say every layer of bureaucracy should be preserved forever. Pruning genuinely redundant reporting lines still makes sense. The real question organisations need to ask before the next round of cuts is: Are we eliminating managers based on what they used to do, or on the critical work they’re doing now and will need to do as AI becomes more embedded?

If the answer leans on the former, you’re treating managers as mere overhead—​a cost centre to be trimmed. The data suggests a different story: managers are the backbone that turns lofty AI ambitions into concrete outcomes on the ground. Assuming AI will auto‑pilot its own adoption, or that value will magically emerge from unguided individual effort, is a productivity gamble the numbers simply don’t back.

In practice, this means investing in managerial capabilities around AI literacy, psychological safety, and experiment‑driven leadership. It also means re‑thinking the phrase “flatten the org” to something more nuanced—​perhaps “optimize the infrastructure so every layer adds clear, measurable value.” When leaders view management as infrastructure rather than overhead, they’re more likely to keep the right people in place, equip them with the right tools, and ultimately unlock the full promise of AI for their teams.

So before you start slashing the next batch of management positions, pause and ask: will the AI initiatives you’ve funded still deliver without the people who make them work?